When validating Dataplane resource, make sure that all inbound and outbound interfaces are unique and don't overshadow one another

[yskopets]

Summary

• When validating Dataplane resource, make sure that all inbound and outbound interfaces are unique and don't overshadow one another
  • it's invalid for an inbound listener to be bound to 0.0.0.0 interface (control plane cannot use this IP in configuration of an Endpoint that represents this Dataplane)
  • every inbound listener must have a unique interface, e.g.
    • it's invalid to have two inbound listeners bound to the same 192.168.0.1:8080 interface
  • every outbound listener must have a unique interface, e.g.
    • it's invalid to have two outbound listeners bound to the same :8080 interface
    • it's invalid to have two outbound listeners where one is bound to :8080 interface and another is bound to 0.0.0.0:8080 interface
  • outbound interfaces must not overshadow inbound interfaces, e.g.
    • if a Dataplane has an inbound interface 192.168.0.1:80:8080 it would be invalid to have the following outbound interfaces:
      • :8080 or 0.0.0.0:8080 (both overshadow application port 8080)
      • 192.168.0.1:80 or 0.0.0.0:80 (both overshadow inbound dataplane port 80)
  • prometheus endpoint (configured via metrics.prometheus) must not overshadow inbound interfaces, outbound interfaces, original application
    • we're already checking this at the moment when we generate Envoy config (because Mesh-wide Prometheus settings might be updated after a Dataplane has been created/updated)
    • additionally, we need to check it at Dataplane modification time as well

[lahabana]

#2799 also suggests an annotation for precedence in case of Clashes which probably makes sense.
In the absence of precedence we should have a total order of services for k8s (headless > clusterIP > alphanumeric).

[lahabana]

Note the slightly related ignore annotation which is the opposite of the primary service exposed in #2799

[github-actions]

This issue was inactive for 30 days it will be reviewed in the next triage meeting and might be closed.
If you think this issue is still relevant please comment on it promptly or attend the next triage meeting.

[github-actions]

This issue was inactive for 30 days it will be reviewed in the next triage meeting and might be closed.
If you think this issue is still relevant please comment on it promptly or attend the next triage meeting.

[github-actions]

This issue was inactive for 30 days it will be reviewed in the next triage meeting and might be closed.
If you think this issue is still relevant please comment on it promptly or attend the next triage meeting.

[github-actions]

Automatically closing the issue due to having one of the "closed state label".

[lahabana]

Triage: Users don't seem to be hitting this so please re-open if you are hitting it :)